Assessing the Complementarities of MALDI and ESI for
Protein Identification in Complex Mixtures
Comparing the AB SCIEX TOF/TOF™ 5800 System with ESI Orbitrap using ProteinPilot™ Software
Sean Seymour1, Aaron Booy2, Rebekah Gundry3, Jennifer Van Eyk3, Christie Hunter1
AB SCIEX, USA, 2AB SCIEX, Canada, 3The Johns Hopkins University, USA
1
There are two dominant ‘soft’ ionization strategies in proteomics
today for biomarker discovery and protein identification in
complex mixtures. On-line LCMS using electrospray ionization
(ESI) is commonly used with mass spectrometers such as ion
traps, quadrupole time of flight, and Orbitrap instruments. Matrix
Assisted Laser Desorption Ionization (MALDI) is the other main
soft ionization technique and is normally coupled with Time of
Flight (TOF) detection systems. With the advent of LC MALDI1, it
has also become widely used in off-line LCMS strategies. While
LC MALDI has much more flexibility in acquisition strategies due
to the separation of LC from MS analysis, it has traditionally
been speed limited relative to the ESI approach.
With the faster laser repetition rate possible on the TOF/TOF™
5800 System and the intelligent DynamicExit™ Algorithm for
efficient acquisition, the speed of LC MALDI is now 10-15X
faster2 then previous instruments. As LC MALDI begins to
approach the speed of on-line ESI strategies, it is worthwhile
investigating the complementarities of the techniques to
understand how to best leverage the strengths of each and get
the most information from complex proteomics samples.
To do the study properly, it was important to assess both the
differences between instruments and the differences within
multiple acquisitions on each instrument. The strategy used was
to measure three replicates of a complex biological sample
(E.coli cell lysate) at three different sample loadings on column
and assess the differences. Reverse database searching with
ProteinPilot™ Software 3.0 was used to measure the numbers of
proteins and peptides detected at a fixed false discovery rate
(FDR) from each instrument.
Figure 1. Summary of ESI and MALDI Acquisitions – Protein Level. Proteins identified at 1% global FDR for each of the three replicate injections at
each sample loading is shown. Replicate acquisitions were performed on both the ESI-Orbitrap platform (red) and the TOF/TOF™ 5800 System (blue) to
enable proper comparison of these orthogonal acquisitions strategies.
p1
Figure 2. Summary of ESI and MALDI Acquisitions – Peptide Level. Peptides identified at 1% global FDR for each of the three replicate injections at
each sample loading is shown. Replicate acquisitions were performed on both the ESI-Orbitrap platform and the TOF/TOF™ 5800 System to enable
proper comparison of these orthogonal acquisitions strategies.
Materials and Methods
Sample Preparation: E.coli lysate was re-suspended in 50%
trifluoroethanol/50 mM ammonium biocarbonate buffer at 1.35
mg/mL (by weight). After heat denaturation, reduction
(dithiothreitol) and alkylation (iodoacetamide) were performed,
the sample was digested with trypsin overnight at 37 °C. Sample
was dried using vacuum centrifugation, then re-suspended at 1
mg/mL in 2% acetonitrile / 0.1% formic acid. Sample loadings of
20, 100 and 500 ngs were run in three replicates for each
loading.
LC MALDI Acquisition: Peptides were separated on a
Monolithic CapRod column (100 µm ID x 15cm, Merck,
Germany) with a 90 minute linear gradient of 5-35% acetonitrile
at 2 µL/min (in 0.1% TFA) using the Tempo™ LC MALDI
Spotting System. 720 spots were collected per LC run with
online mixing of α-cyano-4-hydroxycinnamic acid matrix
(Agilent). MS/MS spectra were acquired on the 25 most intense
precursors per spot using a laser frequency of 1000 Hz and
DynamicExit™ Algorithm on the TOF/TOF™ 5800 System (for
more acquisition details2).
LC ESI Acquisition: Peptides were separated on a self-packed
column (75 µm x 10cm, Magic C18 AQ (Michrom) in a PicoTip
emitter (New Objective)) using a 90 min linear gradient of 5-35%
acetonitrile at 300 nL/min (Agilent 1200 LC system). The eluent
was directed into an LTQ Orbitrap (ThermoFisherScientific). MS
scans were acquired in the Orbitrap and the top 7 precursor ions
above a threshold of 8000 counts were selected for MS/MS in
the LTQ.Data acquisition parameters were optimized according
to the vendor’s recommendations3, and included monoisotopic
precursor selection, 2+, 3+ charge state selection, and a
dynamic exclusion repeat count of 1.
Database searching: Each dataset was searched with
ProteinPilot™ Software 3.0 against a database of all E.coli
proteins in UniProt as of Jan. 27, 2009 (4379) plus 77 common
contaminant proteins using the Paragon™ Algorithm in thorough
mode with biological modifications and substitutions4. False
discovery rate (FDR) analysis was done by on-the-fly analysis of
the reversed sequences using the embedded PSPEP tool5. For
processing of the Orbitrap data, an mgf file was generated using
ReAdw4Mascot2.exe, a NIST derivative of the ISB tool, ReAdW.
Protein and Peptide ID Results
Three different sample loadings of E.coli were analyzed by ESI
Orbitrap and LC MALDI 5800 system and compared using the
protein and peptide identifications at the 1% global FDR. The
processing of the three data acquisition sets is summarized in
Figure 1 and 2. At the protein level (Figure 1), the results at the
20 ng level were quite comparable. At the higher sample loading
of 100 and 500 ngs, a higher number of proteins were routinely
seen using the LC MALDI strategy. The same observation was
seen at the peptide level (Figure 2), with the LC MALDI approach
finding ~2x more peptides at the highest loading amounts.
p2
Figure 3. Intersection of Proteins and Peptides Identified by ESI vs. MALDI. Left: The best replicate acquisitions from the 500 ng sample loading
experiments were compared to assess the intersection between ESI and MALDI strategies, at the protein level (1% global FDR) and the peptide level
(5% local FDR). Right: Correlating the Unused Prot Score (UPS) per protein between instruments clearly shows that more peptide identifications using
the 5800 system yields higher UPS and therefore higher protein confidence.
Protein and Peptide Level Intersection
Using the 500 ng sample loading level, the intersection between
the best ESI-Orbitrap acquisition was compared to the best LC
MALDI 5800 acquisition at the protein and peptide level (Figure
3, left), using the 1% global FDR values. Out of a total of 778
proteins identified, 466 proteins (60%) were shared with the
5800 system finding about 30% more proteins. Even more
significant differences were observed when comparing the
peptide level intersection, where only 25% of peptides were
shared between acquisitions. Here, LC MALDI on the 5800
system found about 108% more peptides than the ESI-Orbitrap
system, meaning more proteins and more sequence coverage
for the detected proteins.
The value of higher numbers of peptides for the purpose of
protein identification can be determined by aligning the protein
level identifications and comparing the Unused Prot Score (UPS
- a measure of protein confidence in ProteinPilot™ Software)
from each acquisition strategy. As a rough approximation, each 2
units of UPS correspond to one high confidence peptide
sequence segment (excluding multiple modification states). The
additional peptide coverage by LC MALDI clearly translates into
more confident protein identifications (Figure 3, right, log scaled)
as seen by the higher proportion of data above the diagonal.
Assessing the Peptide Level Intersection
A comparison using only one replicate is insufficient and
provides only limited insight. It is important to assess both intraand inter-instrument reproducibility. The degree of similarity
between each pair of replicates – both within and across
instruments – was calculated as the ratio of the intersection to
the union of identifications for the replicate pair (Figure 4). The
average intersection at the peptide level between replicates on
the TOF/TOF™ 5800 system at the 500 ng loading was high at
64% (Figure 4, Samples 1-3). The average intersection between
the Orbitrap replicates was also high at 61% (Samples 46).However, if the peptide comparison is done between
platforms, the average intersection decreases to only 25%
(Figure 4, red boxes). Comparison of replicates within each
instrument clearly show more similarity than comparisons of
replicates from different instruments.The source of this difference
is an interesting investigation.
Figure 4. Peptide-Level Intersection between All Acquisitions at 500 ng
Sample Loading Level. Each cell shows the fraction of all peptides found
in each pair of replicates that are common to both. Formatting from red to
green indicates increasing similarity.
p3
Figure 5. Differences in Peptide Detection by ESI and MALDI. Of the peptides detected at 500ng loading, the proportion of each terminating in Lys,
Arg or other residues was determined (left). MALDI finds a higher proportion of Arginine-terminating peptides than ESI. The peptide populations found by
a single technique (right) shows a more distinctive trend. Because of the bias of MALDI towards finding more Arginine-terminating peptides, there were
many more Arg peptides found only in the MALDI strategy, and Lysine-terminating peptides in the ESI strategy.
Differences in Peptide Populations
Previous work has suggested that there is a slight bias towards
specific types of peptides detected between ESI and MALDI
experiments. However, without proper assessment of the
random nature of acquisition, no real trends can be assessed.
Here, using three replicate acquisitions as the basis for our
analysis, we can assess real acquisition based differences. As
shown in Figure 5, there are higher proportions of Arginine (Arg)terminating peptides detected in LC MALDI and this is consistent
across replicates. In the ESI acquisition, the distribution of
peptides between Lysine (Lys) and Arg is roughly equal.
Therefore, when the unique peptides found by only LC MALDI
were analyzed, there was a very high proportion of Arg peptides
found (~67%) vs only ~33% by ESI (Figure 5, right).
The mass of the peptides found were also analyzed, to
understand if there were any real differences in peptide size.
Figure 6 shows the distribution of precursors found across all
replicates. No strong differences are seen between acquisition
strategies, although there are small differences that appear
consistent across replicates.
Figure 6. Mass Distributions of Replicates. The analysis of mass
distributions reveals only small but consistent differences in the masses
of peptides acquired by each instrument.
p4
Conclusions
By analyzing both the intra- and inter-instrument acquisition
replicates, a much clearer picture into the complementary nature
of ESI and MALDI can be ascertained. The LC MALDI using the
AB SCIEX TOF/TOF™ 5800 system reproducibly obtains more
proteins and peptides per acquisition over the ESI-Orbitrap
acquisition at the higher samples loadings. At lower amounts on
column, the acquisition results are comparable.
The analysis of the peptide intersection demonstrates that
distinct peptide populations are detected when using LC MALDI
vs. ESI acquisition (Figure 3). This suggests that a
complementary strategy combining both ESI and MALDI analysis
could provide deeper proteome coverage. Consideration of
combinations of replicates (Figure 7, 500ng) shows that multiple
replicate acquisitions on an ESI-Orbitrap does not equal the
peptide or protein yield from a single LC MALDI run. This
comparison is of interest because of the acquisition time
consideration, 3 LCMS runs takes ~7.5 hours of instrument time
and 1 LC MALDI run takes ~7 hours2.
Another interesting comparison is that the combination of two
5800 system runs is nearly equivalent to a combination of 1 ESIOrbitrap run and 1 LC MALDI run. Since one ESI and one
MALDI run can be done in less time and the differences in the
peptide populations detected, the combined use of these two
platforms might be a valuable strategy for deep proteome
coverage.
References
Figure 7. Relative Benefits of Repeat Acquisitions. The plot of the
ESI-Orbitrap data (red) and the LC MALDI 5800 system data (blue) at
both the protein (top) and peptide level (bottom) demonstrates similar
relative gains from increasing replicates although the absolute numbers
significantly favor the 5800 system. The purple line assesses the
scenario of combining equal numbers of replicates from ESI and MALDI
acquisitions. For example, the combination of 1 5800 system run with 1
Orbitrap run (purple point at 2 replicates) shows more peptide
identifications over a single 5800 system run (blue point at 1 replicate).
1.
Hattan et al. J. Prot. Res. (2005) 4, 1931-1941.
2.
Effects of Intelligent Data Acquisition and Fast Laser Speed
on Analysis of Complex Protein Digests. AB SCIEX
Technical Note 0921810-01.
3.
Advantage of the LTQ Orbitrap for Protein Identification in
Complex Digests, ThermoFisher Scientific App note: 386,
2007.
4.
Shilov IV et al. (2007) Mol. Cell Prot. 6.9, 1638.
5.
Tang WH et al. (2008) J Prot. Res. 7, 3661.
For Research Use Only. Not for use in diagnostic procedures.
© 2010 AB SCIEX. The trademarks mentioned herein are the property of AB Sciex Pte. Ltd. or their respective owners. AB SCIEX™ is being used under license.
Publication number: 0922010-01
p5